Read on Twitter
Seeing an interesting question coming up here - should we just give as many people as possible dose 1 of the Pfizer vaccine and not worry so much about dose 2? Let's do some stats. Thread/
The benefit of this approach is obvious. Vaccines will be in short supply for a while - if we go to a one-dose protocol (for now) we can vaccinate twice as many people. 2/
If one dose were as good as two doses then this would be a no-brainer. But is it? Let's look at the data 3/
This is the cumulative infections chart @Pfizer submitted to FDA. Impressive overall results. But we need to focus on a specific window in time (circled) from 7 days after dose 1 (when immunity would set in) to seven days after dose 2 (which occurred at d21 at the earliest). 4/
This is ALL the data we have to suggest what might happen with a 1 dose vaccine. What do we see? Counting by the dots in that image I get
12 cases in vax group
22 cases in placebo
This translates into an efficacy of 45%... not great.
5/
12 cases in vax group
22 cases in placebo
This translates into an efficacy of 45%... not great.
5/
What if we use cases from d10-d28?
By my eye that works out to:
7 cases in vax
18 cases in placebo.
That's an efficacy of 61%. Better. But there's not much data here. The confidence interval is 7% - 84%! 6/
By my eye that works out to:
7 cases in vax
18 cases in placebo.
That's an efficacy of 61%. Better. But there's not much data here. The confidence interval is 7% - 84%! 6/
If we extend out beyond 28 days, sure it keeps looking better. But at that point we don't know if the increased benefit is due to dose 2 or not. That data is not helpful. 7/
So yes, there is a sign that one dose is beneficial - but it rests on a paper-thin amount of data. It might provide as little as 7% protection, or as much as 84%. We just don't know. 8/
I can't recommend a policy of just giving everyone one dose based on this data. There may be other data I am not aware of, though. This absolutely calls for a randomized trial of 1 versus 2 doses though as @thehowie has suggested. 9/
I'd love to see such a trial. For now, putting stock in the one dose approach appears to be motivated reasoning. 10/10.
Update:
Some folks have correctly pointed out that each dot is not a case. We can infer cases from the no at risk. This makes case for one dose a bit stronger, but confidence intervals are still wide: https://twitter.com/fperrywilson/status/1337153301423517697
Some folks have correctly pointed out that each dot is not a case. We can infer cases from the no at risk. This makes case for one dose a bit stronger, but confidence intervals are still wide: https://twitter.com/fperrywilson/status/1337153301423517697
